AI 2

Is MR doing AI wrong?

Artificial Intelligence. Machine Learning. It’s the future. 

Or is it?

The Market Research (MR) sector is, like many others, beginning to embrace new and emerging technology and software, with all of the jargon and promise they bring. But is it fit for purpose? Or is a square peg being brute forced into a round hole, simply because it’s the latest thing?

It’s a conversation we had internally at Insightflow, and it prompted us to ask questions of various experts and MR practitioners. There’s a concern in some corners of the internet that a reliance on the use of AI in MR could turn researchers into glorified accountants. 

As one respected researcher put it: “I’m pretty much seeing the market as smart companies selling mediocre software to dumb buyers.”

So what can AI actually do for MR? At the moment, it would appear to be heavy lifting and legwork in quantitative and not too much in qualitative. 

Ascribe meaning

If your research is a matter of finding the number of times X is mentioned, then AI will massively simplify that process and produce useful data you can then interpret and gain insights from. However, if your research is qual and requires nuance, then you may have issues. As Adrian Rhodes of AR Research says, “Human intervention can ascribe meaning.”

Ben Smithwell, another highly respected researcher, is concerned that a tendency by businesses to place more faith in quantitative research and numerical data may see MR pressured to use AI because it appears to be more scientific. He saw it happening in marketing. “Marketers have turned into glorified accountants – the same thing is happening for researchers.”

So how can MR best apply AI technology? One area where AI should definitely be able to help in research is in making the process of data collection more effective and thereby freeing up researchers to do the human aspects of the work.

Enable the humans

“I suppose we have to define ‘a tool’ here – I think it is something that amplifies the skills and capabilities of the user. We need to see AI as a tool that enables this,” Smithwell says. 

For those involved in the development of AI tools, however, the picture is rosier. Jason More, of Morescode and formerly of Chattering Monkey, says: “AI can now do both quant and qual, it allows users to stay close to the data (of course) but is also now allowing users to get closer to the ‘story’. I think AI is starting to merge qual and quant, certainly to reduce the conceptual distance between them. AI allows qual to drill into larger data sets.”

The key stumbling block appears to be in expectation from both MR and business clients. “…cost squeeze in research AI has led to a low level of satisfaction. People didn’t understand what it could deliver well enough to be able to specify what it should deliver,” More says. 

“At Chattering Monkey we were often asked to deliver projects that were outside the core product set. We could do that and often did, but to me it seemed to be an indication of customers buying AI for the sake of being able to report that ‘they are using AI’ rather than because of well thought through use cases.”

Incitements to excitement

It’s clear that if you have large amounts of data to sift, then AI tools can simplify that process massively. But there are pitfalls for research. For example, Adrian Rhodes cites the use of speech recognition software, an area where AI can be very good, but still presents errors. Sentiment Analysis via AI transcript is problematic as a result. He likens the potential risks of false positives as being akin to the early days of sat nav. 

For those at the sharp end, like Smithwell, the future uses of AI and the potential for really great tools present the most interesting possibilities. 

“Pieces of data should be seen as incitements to excitement” he says. The implication is that the excitement comes from what humans can do with great data and in turn it’s great tools that can produce the input pieces of data.

“I’d love to be in a position where, once I have made an assumption that may – or may not be right (and it’s based on ‘dog work’, I can check the clusters and big data via AI.”

As Brian Livell of Big Sofa Technology says, “Artificial Intelligence will have a positive impact on the insights industry – improving data quality, consumer engagement, and strategic decisions.” 

But he also notes, “However, technology-only solutions will not be able to meet all of the demands within the industry. Companies that can combine AI-based technology with expert consultancy will have significant advantages in the industry over those that fail to embrace AI as a driving force.”

The limits on the use of AI today appear to be down to several things, not least the tools presently on offer, with several applications where AI is a solution looking for a problem. The other issue is in the data offered to AI tools and the fact that most companies fail to appreciate just how powerful the use of AI could be, if applied to all of a company’s data. As Jason More notes:  “Where we are at with AI is this:  It has the ability to do a lot of things, most of what we might ask of it. What does not exist is the data (in sufficient quantities) in the right place. This may remain a problem with research where the data sets are often not huge. But businesses generate a lot of data across a whole variety of different channels and there is the ability to use AI to bring this together in meaningful ways.”

Currently, as Adrian Rhodes puts it, “AI is a means, not an end.” But that could change, if AI developers create the right tools and MR companies understand that more data means more results.

Related Articles